The asymptotic distribution of the characteristic roots and (normalized) vectors of a sample covariance matrix is given when the observations are from a multivariate normal distribution whose covariance matrix has characteristic roots of arbitrary multiplicity. The elements of each characteristic vector are the coefficients of a principal component (with sum of squares of coefficients being unity), and the corresponding characteristic root is the variance of the principal component. Tests of hypotheses of equality of population roots are treated, and confidence intervals for assumed equal roots are given; these are useful in assessing the importance of principal components. A similar study for correlation matrices is considered.
%0 Journal Article
%1 anderson1963asymptotic
%A Anderson, T. W.
%D 1963
%J Ann. Math. Statist.
%K PCA covariance_estimation covariance_matrix sample_covariance_matrix spectral_theory
%P 122--148
%T Asymptotic theory for principal component analysis
%V 34
%X The asymptotic distribution of the characteristic roots and (normalized) vectors of a sample covariance matrix is given when the observations are from a multivariate normal distribution whose covariance matrix has characteristic roots of arbitrary multiplicity. The elements of each characteristic vector are the coefficients of a principal component (with sum of squares of coefficients being unity), and the corresponding characteristic root is the variance of the principal component. Tests of hypotheses of equality of population roots are treated, and confidence intervals for assumed equal roots are given; these are useful in assessing the importance of principal components. A similar study for correlation matrices is considered.
@article{anderson1963asymptotic,
abstract = {The asymptotic distribution of the characteristic roots and (normalized) vectors of a sample covariance matrix is given when the observations are from a multivariate normal distribution whose covariance matrix has characteristic roots of arbitrary multiplicity. The elements of each characteristic vector are the coefficients of a principal component (with sum of squares of coefficients being unity), and the corresponding characteristic root is the variance of the principal component. Tests of hypotheses of equality of population roots are treated, and confidence intervals for assumed equal roots are given; these are useful in assessing the importance of principal components. A similar study for correlation matrices is considered.},
added-at = {2014-07-03T18:40:50.000+0200},
author = {Anderson, T. W.},
biburl = {https://www.bibsonomy.org/bibtex/29c2f7f65c52552242a296a3e3ac74a65/peter.ralph},
fjournal = {Annals of Mathematical Statistics},
interhash = {73d98c6b5db529191534c797d28f5e25},
intrahash = {9c2f7f65c52552242a296a3e3ac74a65},
issn = {0003-4851},
journal = {Ann. Math. Statist.},
keywords = {PCA covariance_estimation covariance_matrix sample_covariance_matrix spectral_theory},
mrclass = {62.40},
mrnumber = {0145620 (26 \#3149)},
mrreviewer = {R. L. Plackett},
pages = {122--148},
timestamp = {2014-07-03T18:40:50.000+0200},
title = {Asymptotic theory for principal component analysis},
volume = 34,
year = 1963
}